Introduction

Despite a worldwide decline in incidence, gastric cancer (GC) still is the second most common cause of cancer-related mortality.1 It is well accepted that Helicobacter pylori infection is a key risk factor for GC. However, only a small fraction of the infected people develop GC or its precursors.2, 3, 4 Such clinical diversity suggests that factors other than bacterial infection alone determine gastric carcinogenesis. Apart from virulence factors of the pathogen and other environmental and lifestyle risk factors, host genetic susceptibility is also likely to contribute.5, 6, 7 In recent years, host genetic polymorphisms involved in inflammatory response, carcinogen metabolism, antioxidant protection, mucosal protection and cell proliferation regulation have been widely studied as potential biomarkers to predict GC risk. However, the findings are frequently heterogeneous.8, 9, 10 In this article, we provide a systematic review of studies addressing the association of cell proliferation-related polymorphisms with GC susceptibility. For the most widely studied polymorphisms, we also quantitatively summarized associations with GC using meta-analysis.

Materials and methods

Literature search

Studies investigating associations of host genetic polymorphisms and GC risk were identified by searching for articles in the MEDLINE database and Chinese BioMedical Literature Database. Articles published until 15 September 2008 were considered. Different combinations of the key words ‘gastric cancer’, ‘stomach cancer’, ‘polymorphism(s)’ and ‘susceptibility’ were used to screen for potentially relevant studies. Additional studies were also identified using cross-referencing.

Inclusion and exclusion criteria

Polymorphisms related to GC risk were divided into several groups according to their biological roles: mucosal protection and inflammation response, carcinogen metabolism, oxidative damage and DNA reparation, tumor invasion (cell adhesion and angiogenesis) and the regulation of cell proliferation. Cell proliferation-related genetic polymorphisms were selected for this review. Case–control or cohort studies presenting original data on associations between the genetic polymorphisms and GC were included. If the effect of a polymorphism was reported in duplicate, the article published in English or published earlier was included. Exclusion criteria include: (i) articles not in English or Chinese; (ii) review articles; (iii) articles which were not cases–control or cohort studies addressing GC susceptibility; (iv) articles focusing on polymorphisms other than cell proliferation-related.

Data extraction and statistical analysis

Data extraction and analysis followed standard methods for systematic review and meta-analysis11, 12 as described elsewhere.13 Briefly, for all studies, we extracted the following data from the original publications: first author and year of publication; genes and relevant polymorphisms; characteristics of the study design and the study population, association according to tumor location and histological types, case–control matching criteria and covariates controlled for (the latter are presented in Supplementary Tables only). Adjusted odds ratios (ORs) were extracted from the studies where available and included in tabular presentation. For some studies, crude ORs had to be calculated from the reported frequencies of genotype by disease status.

Meta-analyses were carried out using Comprehensive Meta-Analysis (V2.0, Biostat, Englewood, NJ, USA) for the most widely studied polymorphisms, which were evaluated in at least five studies. Dominant models were used to represent the effect of the polymorphism, and corresponding ORs (adjusted ORs were used where available) from the included studies were summarized. Random effects models were used in meta-analysis, taking into account the possibility of heterogeneity between studies, which was evaluated with the Q test (P<0.10 was considered indicative of statistically significant heterogeneity) and the I2 statistic (values of 25, 50 and 75% are considered to represent low, medium and high heterogeneity, respectively). The Begg rank correlation method and the Egger weighted regression method were used to statistically assess publication bias (P<0.05 was considered indicative of statistically significant publication bias). Stratified analyses were conducted according to the ethnicity of the study population (Asian or Caucasian) and histological type of GC (Lauren's classification: diffuse type or intestinal type).

Results

The literature search identified 61 original articles14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74 on associations between cell proliferation-related genetic polymorphisms and GC (flow diagram of study identification shown in Figure 1). Of these, six articles69, 70, 71, 72, 73, 74 were excluded from this review because of duplication reporting of results. From 55 included studies, 54 were hospital-based (n=46) or population-based (n=8) case–control studies, and only one was a cohort study-based nested case–control study.34 Forty polymorphisms in 27 genes were assessed in included studies and 23 of them were significantly related to GC in at least one study. On the basis of their biological roles, the involved polymorphisms were categorized in two groups: cell cycle and apoptosis regulators (Table 1 and Supplementary Table 1) and cell growth factors-related (Table 2 and Supplementary Table 2) polymorphisms.

Figure 1
figure 1

Flow diagram of study identification (until 15 September 2008).

Full size image
Table 1 Association between cell cycle and apoptosis regulators gene polymorphisms and gastric cancer susceptibility
Full size table
Table 2 Association between growth factors gene polymorphisms and gastric cancer susceptibility
Full size table

There were 43 studies focusing on the relationships between cell cycle regulators-related polymorphisms and GC susceptibility (Table 1). Inconsistent associations were observed for tumor protein p53 (TP53) Arg73Pro, the most widely studied polymorphism. Meta-analysis of the 14 included studies showed that, overall, the Pro allele at codon 72 of TP53 was not significantly associated with GC (OR, 1.08; 95% confidence interval (CI), 0.91–1.27) (Figure 2).14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 25, 26, 27, 28 However, when stratified by ethnicity and histological type of GC, we found that among Asians, the Pro allele acted as risk factor of GC and this association was particularly pronounced in diffuse GC (OR, 1.44; 95% CI, 1.04–1.99) but absent in intestinal GC (OR, 1.07; 95% CI, 0.64–1.80). No substantial difference was observed between hospital-based (OR, 1.26; 95% CI, 1.03–1.53) and population-based studies (OR, 1.13; 95% CI, 0.87–1.47) among Asians. Among Caucasians, the Pro allele was associated with a reduced risk of intestinal GC (OR, 0.56; 95% CI, 0.36–0.89), and no significant association was found with diffuse GC (OR, 0.74; 95% CI, 0.48–1.16). However, only six out of the 14 included studies reported associations by histological type of GC (four in Asian populations15, 20, 26, 27 and two in the Caucasian populations16, 25), and therefore the summarized results have to be interpreted with caution. One study on this polymorphism shown in Table 1 was excluded from the meta-analysis because the genotype frequencies and ORs were not reported by the original article.24 No evident publication bias was observed as assessed by Begg rank correlation method (P=0.83) and Egger weighted regression method (P=0.82).

Figure 2
figure 2

Meta-analysis of TP53 72Pro and gastric cancer susceptibility.

Full size image

Polymorphisms of functional regulators of TP53, TP53BP2 (tumor protein p53 binding protein 2) and MDM2 were also found to be related to the development of GC in studies from Korea and Japan, respectively.29, 30 An SNP (exon 2 4G>A) in TP73, which encodes a homology of p53, showed significant association in Chinese32 but not in Japanese.14 Four studies were carried out on the 870G>A variant in Cyclin D;33, 34, 35, 36 a significantly reduced risk of GC was observed for AA genotype in one study36 carried out in the Chinese population. H-RAS was reported to stabilize p21 by promoting the formation of p21–cyclin D1 complexes that prevent subsequent degradation; Harvey retrovirus-associated DNA sequences (H-RAS) 81C allele was associated with an increased risk of GC (OR, 3.7; 95% CI, 2.2–6.0) in a study from China.38

Meta-analysis showed inconsistent results for the S allele of an EcoRI polymorphism in the second intron of L-myc (MYCL1) (Figure 3). Of the four included Asian studies, one showed a significant increase in risk, the other three presented non-significant associations in opposite directions.39, 40, 42, 45 Diverse results were also found in the two studies involving Caucasians.43, 44 One study on this polymorphism shown in Table 1 was excluded from meta-analysis because the allele frequencies were not reported by the original article.41 The results of Begg rank correlation analysis (P=0.13) and Egger weighted regression analysis (P=0.49) did not indicate significant publication bias.

Figure 3
figure 3

Meta-analysis of L-myc S allele and gastric cancer susceptibility.

Full size image

Genotypes inhibiting apoptosis were reported to increase GC risk. For example, Survivin -31C allele was associated with an increased risk compared with GG genotype in studies from China.46, 47 Gene variants in transcription factors participating in cell cycle regulation, such as peroxisome proliferator-activated receptor γ (PPAR-γ), eukaryotic translation release factor 3 , krüppel-like factor 6 and RUNX3, were also identified as potential predictors of GC.49, 50, 51, 52, 53, 54

Recently, polymorphisms related with growth factors were assessed as potential risk markers of GC (Table 2). At 5′ UTR in epidermal growth factor (EGF) gene, 61G carrier was associated with reduced GC risk in the Asian studies from Japan and China.56, 57, 58 Ile655Val variant of HER2, a member of the EGF receptor family, showed a significant association with GC.59 TGFB1-509T and TGFBR2-875A were reported determining high levels of transforming growth factor β1 (TGF-β1) and TGF-β receptor 2 (TGF-βR2), respectively. Decreased risks of GC were observed for these two alleles in one study from China.60 Li ZQ et al64 studied polymorphisms of insulin-like growth factors (IGF)-related genes. No association was found between GC risk and SNPs of insulin (INS), IGF-II, insulin-like growth factor binding protein 1 (IGFBP1) and IGF-1 receptor (IGF1R) genes. However, a significant association was observed for IGFBP3 -202A>C in a British population65 and for Gly32Ala in a Chinese population.64 Pepsinogen C (PGC) insertion/deletion variant between exons 7 and 8 was significantly associated with GC risk in studies carried out among the Chinese and Portuguese populations in reverse directions.67, 68

Discussion

This systematic review addressed the association of cell proliferation-related genetic polymorphisms with GC reported up to 15 September 2008. Twenty-three polymorphisms significantly related to GC in at least one published study were identified, which suggests that polymorphisms in genes implicated in cell proliferation could be candidate biomarkers of GC risk.

Cell cycle and apoptosis regulators, which are directly involved in the initiation of cellular malignant proliferation, have long been preferred targets as cancer risk markers.75 Our analysis regarding polymorphisms in TP53 are consistent with and extend findings from a recent meta-analysis focusing on the Arg73Pro polymorphism.76 Our meta-analysis, which included two additional studies on this polymorphism,19, 21 confirmed associations to vary by population and histological type of GC. Significant heterogeneity was observed among all the included 14 studies on TP53 72Pro, with no evidence of an overall association with GC risk. When stratified by ethnicity, however, studies included in subgroup analyses displayed better homogeneity, with an indication of an increased risk among the Arg allele carriers in Asians and a tentatively reduced risk among Caucasians. Such inverse associations might be explained by differing environmental factors, which act jointly with either apoptotic or DNA repair machinery, respectively. Environmental risk factors might also determine the mutational spectrum of TP53, which was also shown to contribute to the functional consequences of the Pro72Arg polymorphism.77 Alternatively, a different degree of linkage disequilibrium in different ethnicities of this variant with another functional variant might underlie the findings. One candidate polymorphism could be the 16-bp duplication in intron 3 that has previously been associated with reduced TP53 mRNA.78

Heterogeneous results were found for the other two most widely studied polymorphisms in cell cycle-related genes, L-myc EcoRI polymorphism39, 40, 41, 42 and p21 Arg31Ser,18, 19, 21, 37 even in the same ethnicity. Such an inconsistency may be partly explained by study design and the interaction with non-genetic risk factors, such as H. pylori infection and dietary factors, which strongly vary between populations. So, besides stratification by ethnicity and different type of GC, larger population-based studies with careful ascertainment of ‘lifestyle’ and ‘environmental’ factors are essential to fully understand the role of host genetic susceptibility.

A consistently increased risk was observed for PPAR-γ 12Ala carriers in the Chinese, Japanese and Indian populations.51, 52, 53 PPAR-γ is a member of the nuclear hormone receptor family that plays an important role in cell differentiation and regulation of metabolism. A potential interplay between PPAR-γ Pro12Ala polymorphism and H. pylori infection was observed in the development of GC.51, 53 It was also suggested that this polymorphism may be associated with gastric mucosal atrophy in H. pylori-infected patients, thereby increasing the risk of GC.52 Further studies on different ethnic groups are needed to confirm the observed association and to clarify the role of PPAR-γ during gastric carcinogenesis.

Polymorphisms determining higher level of growth factors and related receptors, which are important to tissue repair, were associated with reduced risk of GC. Such associations were observed for EGF 5′ UTR 61G>A,56, 58 TGFB1 -509C>T,60 TGFBR2 -875G>A,60 and IGFBP3 -202A>C65 and Gly32Ala.66 PGC was reported to not only act as a digestive enzyme, but might also be a growth factor during the healing of gastric lesions.79 Genetic polymorphisms in PGC gene determining lower expression were also supposed to contribute to gastric ulcer and GC by failing to prevent disease development.68, 80

Although this review indicates that cell proliferation-related genetic polymorphisms could be candidate biomarkers in GC risk, their overall effect seems to be modest and results were often inconsistent. Our analyses suggest that the inconsistencies may be explained, in part, by differences between the study populations and potentially different effects on different types of GC. In addition, different covariates were considered and controlled in different studies as presented in the Supplementary Tables. Considering H. pylori infection is suggested to be a (close to) necessary condition for development of noncardia gastric cancer,81, 82 the role of genetic polymorphisms may primarily be restricted to some (minor) modulation of the risk in the presence of H. pylori infection. In consideration of the potential misclassification of H. pylori status due to disease-related clearance of infection,83 reports on gene–H. pylori interaction should be interpreted with caution.

There are some limitations to this systematic review that need careful consideration. First, because of the limited information supplied by included studies and the small sample sizes, relevant stratifications (eg, by ethnicity or type of GC) could not be made for many studies. Second, some of the included studies did not mention whether polymorphisms in controls were in Hardy–Weinberg equilibrium. However, no significant deviation from Hardy–Weinberg equilibrium was observed for all included studies based on our own calculations (data not shown), except the study by Xie et al,37 which presented allele frequencies only. Third, owing to the heterogeneity in length and detail of presentation of the included studies, no consistent formal rating of quality of studies was possible. Fourth, observed associations in the meta-analyses were generally weak and may partly reflect false positive results due to multiple testing.

In conclusion, this systematic review suggested that cell proliferation-related genetic polymorphisms could be candidate biomarkers of GC risk, but current evidence for the use for risk stratification is still very limited. Stratifications by ethnicity and GC type seem to be crucial in future studies aiming to clarify the effect of genetic polymorphisms on GC risk.